Move from the first enquiry to a complete client handover.

1. Turn the enquiry into an engagement record.

Record the client, objective, expected deliverable, owner and missing information. AI may propose an email summary, but it must distinguish what the client stated from questions that remain open. An urgent tone is not an agreed deadline.

Assign an identifier before creating folders and tasks. A second email from the same client may concern a different engagement: matching by sender alone is insufficient. Resolve ambiguity before documents from different projects enter the same collection. This is especially useful when several colleagues manage work for one client.

2. Define when the work is ready to start.

For each engagement type, list the required inputs and who must provide them. If a file is missing, the case is waiting for that file rather than generically in progress. Reminders can be prepared or sent under approved conditions, avoiding requests for information already received.

Scroll the table to compare all columns.

3. Automate repeatable preparation.

Useful candidates include project-folder setup, document checklists, status summaries and an open-questions list. AI may prepare a draft from defined sources, with references. It should not fill missing client information with plausible details. A visibly incomplete draft is more useful than a polished document that hides unsupported assumptions.

Keep client contexts separate and limit source collections and permissions to the engagement. Reusable templates should contain fields to complete rather than names, amounts or details left over from previous work. Inspect both content and document properties before automating template copying.

4. Make revisions and scope changes explicit.

A request during delivery may be a correction, a choice between alternatives or an extension to the engagement. Record its type, author, date and decision rather than treating every comment as an immediately authorised task. When scope changes, the owner reassesses timing and responsibilities.

Use an identifiable review version and one agreed location for the current document. An AI comment summary is useful when every point links to the original comment and unresolved questions remain visible. The reviewer needs to distinguish proposed wording from an accepted change.

• Document identifier and review version.
• Comments with an owner and current status.
• Recorded decisions for extensions and out-of-scope requests.
• Explicit approval before the work is labelled final.

5. Complete the handover, not just the document.

Delivery includes the approved version, correct recipients, expected attachments and instructions needed for the next step. If the client must confirm something or the team has a follow-up action, keep it visible. A generated document is not evidence of a completed delivery.

A handover record states what was produced, what remains open, where the material lives and who owns subsequent actions. Archiving, access and retention should follow the firm’s approved rules. Specific professional requirements need review by the relevant responsible people, including differences between markets where applicable.

6. Measure the handoffs that create delays.

For a first pilot, choose a recurring engagement type and compare similar cases. Measure waits for documents, revisions caused by missing inputs, unassigned tasks and effort needed to prepare the final package. These indicators show whether automation simplifies coordination without attributing the quality of professional judgement to a workflow tool.

• One reusable checklist for one engagement type.
• An owner who can review summaries and documents.
• Test cases with ambiguity, revisions and missing inputs.
• A manual way to continue when processing is interrupted.

The examples describe possible procedures; they are not client results or professional advice for a specific engagement.

Practical questions